The solar acoustic oscillations are likely stochastically excited by convective dynamics in the solar photosphere, though few direct observations of individual source events have been made and their detailed characteristics are still unknown. Wave source identification requires measurements that can reliably discriminate the local wave signal from the background convective motions and resonant modal power. This is quite challenging as these ‘noise’ contributions have amplitudes several orders of magnitude greater than the sources and the propagating wave fields they induce. In this paper, we report on a new robust method for the unambiguous identification of acoustic source sites in the photosphere of a MPS/University of Chicago Radiative MHD (MURaM) magnetohydrodynamic simulation of the upper solar convection zone. The method was developed by first utilizing a deep learning algorithm to reliably identify the weak residual high-frequency signature of local acoustic sources, the two-dimensional acoustic Green’s function response of the atmosphere, in Doppler velocity maps and then deciphering what underlies its success. We have diagnosed what the learning algorithm is detecting, mimicked the filter it is applying, and applied the filter directly to the simulated photospheric time series, bypassing the dependence on deep-learning and allowing direct visualization of the local wave pulses that propagate outward from the acoustic source sites. To be effective, the acoustic-source filter thus derived requires high cadence (< 3 seconds) and high spatial resolution (< 50 km) timeseries. Fortuitously, the observational capabilities required to apply the filter to real solar data are just now becoming available with the commissioning of the National Science Foundation’s Daniel K. Inouye Solar Telescope (DKIST). Using the filter developed, we have uncovered previously unknown properties of the acoustic emission process. In the simulation, acoustic events are found to be clustered at mesogranular scales, with peak emission quite deep, about 500 km below the photosphere, and sites of very strong emission can result from the interaction of two supersonic downflows that merge at that depth. We suggest that the method developed, when applied to high-resolution high-cadence observations, such as those forthcoming with Daniel K. Inouye Solar Telescope (DKIST), will have important applications in chromospheric wave-studies and may lead to new investigations in high-resolution local-helioseismology.